Commit graph

29 commits

Author SHA1 Message Date
ehouse
76c0a518f1 change -j to -j3. No code change. 2009-04-14 03:27:32 +00:00
ehouse
641ae18864 Use new dict api to support utf8 faces. 2009-04-05 19:20:22 +00:00
ehouse
ce2420840d a bit of cleanup 2008-11-22 19:43:47 +00:00
ehouse
6e9f5791d6 Fix bug: when using hint limits to restrict first move on empty board
limits were ignored for the vertical search.  Fix is to make
special-case treatment of first move more integrated so special limits
code isn't needed.
2008-11-22 19:32:42 +00:00
ehouse
f749fdd560 Remove all tabs. No code changes. 2008-05-31 03:26:16 +00:00
ehouse
ebad668227 add some consts. 2008-05-06 12:49:37 +00:00
ehouse
1172502b04 If search limits in use, try vertical as well as horizontal initial moves. 2008-03-29 19:40:03 +00:00
ehouse
eeebeca542 substitute macro for common sizeof(x)/sizeof(x[0]) pattern; no generated code change 2007-05-26 14:03:07 +00:00
ehouse
f56b170b2a cleanup. No change in generated code. 2007-01-27 15:07:02 +00:00
ehouse
1838640823 Fix bug where sign extention on 16-bit processors caused too many bits
set in crosschecks.
2006-11-10 08:08:59 +00:00
ehouse
213cfbb050 set crosschecks inline 2006-09-23 15:12:10 +00:00
ehouse
16519a2070 fix bug checking crosschecks with more than 32 tiles. 2006-09-02 05:29:13 +00:00
ehouse
4daabf6fe5 Set -Wunused-parameter for those versions of gcc that support it, and
deal with the output by removing params where possible and elsewhere
by adding XP_UNUSED macro wrapping __attribute__((unused)).  There
should be NO change in function in spite of the large number of files.
2006-08-16 13:44:44 +00:00
ehouse
5d161cdd77 optimize for the case where tiles are < 32. Previous imlementation of
64-tiles support increases engine times by ~5%.  This removes that effect.
I can't measure any slowdown vs. pre-change engines.
2006-08-12 01:59:03 +00:00
ehouse
c41b211735 increase possible number of unique tiles from 32 to 64 in order to
support Slovak which has 40.  This change is preliminary: it seems to
work for both existing and dicts with >32 chars, but it breaks backwards
compatibility: existing saved games will crash this code.  I think the
fix is to deduce the number of bits per tile from the number in the
dictionary being used, but that's not in yet.
2006-08-09 04:56:34 +00:00
ehouse
92485783af update email address in header comments: no code change 2006-01-08 01:25:02 +00:00
ehouse
6728862399 cast to stop warning on wince 2005-06-27 03:56:09 +00:00
ehouse
27b6f310b5 bug: comparison was dropping words with the same score as all those
already cached.
2005-01-25 04:39:55 +00:00
ehouse
aa6e46d05f always set nTilesMax so works when XWFEATURE_SEARCHLIMIT not defined 2004-12-11 04:16:48 +00:00
ehouse
fd75f3324e Pass number of blanks to util_engineStarting 2004-10-07 13:51:03 +00:00
ehouse
38a4170fe0 cleanup; fix for 4-byte nodes on ARM 2004-09-28 01:03:07 +00:00
ehouse
d9efe4258b cleanup for no hintlimit feature case 2004-06-27 04:35:42 +00:00
ehouse
c6c6d38dfc engine is now responsile for asking for and tracking how many tiles in
the tray should be used in hints.  Moved responsibility here from
platform because platform doesn't know when a turn ends, and end of a
turn (engine reset) is the right place to reset the values.
2004-06-25 06:49:49 +00:00
ehouse
7914876294 Change engine to, when given a rect that's a subset of the board,
constrain the search for moves to those that involve filling only
tiles within the rect -- while still using the rest of the board as
context for the moves.  Robot moves use the whole board.
2004-06-15 01:58:09 +00:00
ehouse
4fdf6fe60a new param to figureMoveScore 2004-04-30 08:27:24 +00:00
ehouse
997f0884cc new macros for new dawg node format 2003-12-14 16:59:58 +00:00
ehouse
e5801bd6f6 fix compiler warning 2003-12-13 18:37:43 +00:00
ehouse
1072bf2499 add option to limit number of tiles used by robot 2003-11-30 09:34:40 +00:00
ehouse
5776da0b93 first checkin 2003-11-01 05:35:29 +00:00